static
THREAD_FUNCTION thread_encoding_proc(void *p_data)
{
int ithread = ((ENCODETHREAD_DATA *)p_data)->ithread;
VP8_COMP *cpi = (VP8_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr1);
MB_ROW_COMP *mbri = (MB_ROW_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr2);
ENTROPY_CONTEXT_PLANES mb_row_left_context;
while (1)
{
if (cpi->b_multi_threaded == 0)
break;
if (sem_wait(&cpi->h_event_start_encoding[ithread]) == 0)
{
const int nsync = cpi->mt_sync_range;
VP8_COMMON *cm = &cpi->common;
int mb_row;
MACROBLOCK *x = &mbri->mb;
MACROBLOCKD *xd = &x->e_mbd;
TOKENEXTRA *tp ;
#if CONFIG_REALTIME_ONLY & CONFIG_ONTHEFLY_BITPACKING
TOKENEXTRA *tp_start = cpi->tok + (1 + ithread) * (16 * 24);
const int num_part = (1 << cm->multi_token_partition);
#endif
int *segment_counts = mbri->segment_counts;
int *totalrate = &mbri->totalrate;
if (cpi->b_multi_threaded == 0) /* we're shutting down */
break;
for (mb_row = ithread + 1; mb_row < cm->mb_rows; mb_row += (cpi->encoding_thread_count + 1))
{
int recon_yoffset, recon_uvoffset;
int mb_col;
int ref_fb_idx = cm->lst_fb_idx;
int dst_fb_idx = cm->new_fb_idx;
int recon_y_stride = cm->yv12_fb[ref_fb_idx].y_stride;
int recon_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride;
int map_index = (mb_row * cm->mb_cols);
volatile const int *last_row_current_mb_col;
volatile int *current_mb_col = &cpi->mt_current_mb_col[mb_row];
#if (CONFIG_REALTIME_ONLY & CONFIG_ONTHEFLY_BITPACKING)
vp8_writer *w = &cpi->bc[1 + (mb_row % num_part)];
#else
tp = cpi->tok + (mb_row * (cm->mb_cols * 16 * 24));
cpi->tplist[mb_row].start = tp;
#endif
last_row_current_mb_col = &cpi->mt_current_mb_col[mb_row - 1];
/* reset above block coeffs */
xd->above_context = cm->above_context;
xd->left_context = &mb_row_left_context;
vp8_zero(mb_row_left_context);
xd->up_available = (mb_row != 0);
recon_yoffset = (mb_row * recon_y_stride * 16);
recon_uvoffset = (mb_row * recon_uv_stride * 8);
/* Set the mb activity pointer to the start of the row. */
x->mb_activity_ptr = &cpi->mb_activity_map[map_index];
/* for each macroblock col in image */
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
{
*current_mb_col = mb_col - 1;
if ((mb_col & (nsync - 1)) == 0)
{
while (mb_col > (*last_row_current_mb_col - nsync))
{
x86_pause_hint();
thread_sleep(0);
}
}
#if CONFIG_REALTIME_ONLY & CONFIG_ONTHEFLY_BITPACKING
tp = tp_start;
#endif
/* Distance of Mb to the various image edges.
* These specified to 8th pel as they are always compared
* to values that are in 1/8th pel units
*/
xd->mb_to_left_edge = -((mb_col * 16) << 3);
xd->mb_to_right_edge = ((cm->mb_cols - 1 - mb_col) * 16) << 3;
xd->mb_to_top_edge = -((mb_row * 16) << 3);
xd->mb_to_bottom_edge = ((cm->mb_rows - 1 - mb_row) * 16) << 3;
/* Set up limit values for motion vectors used to prevent
* them extending outside the UMV borders
*/
x->mv_col_min = -((mb_col * 16) + (VP8BORDERINPIXELS - 16));
x->mv_col_max = ((cm->mb_cols - 1 - mb_col) * 16) + (VP8BORDERINPIXELS - 16);
x->mv_row_min = -((mb_row * 16) + (VP8BORDERINPIXELS - 16));
x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16) + (VP8BORDERINPIXELS - 16);
xd->dst.y_buffer = cm->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
xd->dst.u_buffer = cm->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
xd->dst.v_buffer = cm->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
xd->left_available = (mb_col != 0);
x->rddiv = cpi->RDDIV;
x->rdmult = cpi->RDMULT;
/* Copy current mb to a buffer */
vp8_copy_mem16x16(x->src.y_buffer, x->src.y_stride, x->thismb, 16);
if (cpi->oxcf.tuning == VP8_TUNE_SSIM)
vp8_activity_masking(cpi, x);
/* Is segmentation enabled */
/* MB level adjustment to quantizer */
if (xd->segmentation_enabled)
{
/* Code to set segment id in xd->mbmi.segment_id for
* current MB (with range checking)
*/
if (cpi->segmentation_map[map_index + mb_col] <= 3)
xd->mode_info_context->mbmi.segment_id = cpi->segmentation_map[map_index + mb_col];
else
xd->mode_info_context->mbmi.segment_id = 0;
vp8cx_mb_init_quantizer(cpi, x, 1);
}
else
/* Set to Segment 0 by default */
xd->mode_info_context->mbmi.segment_id = 0;
x->active_ptr = cpi->active_map + map_index + mb_col;
if (cm->frame_type == KEY_FRAME)
{
*totalrate += vp8cx_encode_intra_macroblock(cpi, x, &tp);
#ifdef MODE_STATS
y_modes[xd->mbmi.mode] ++;
#endif
}
else
{
*totalrate += vp8cx_encode_inter_macroblock(cpi, x, &tp, recon_yoffset, recon_uvoffset, mb_row, mb_col);
#ifdef MODE_STATS
inter_y_modes[xd->mbmi.mode] ++;
if (xd->mbmi.mode == SPLITMV)
{
int b;
for (b = 0; b < xd->mbmi.partition_count; b++)
{
inter_b_modes[x->partition->bmi[b].mode] ++;
}
}
#endif
/* Special case code for cyclic refresh
* If cyclic update enabled then copy
* xd->mbmi.segment_id; (which may have been updated
* based on mode during
* vp8cx_encode_inter_macroblock()) back into the
* global segmentation map
*/
if (cpi->cyclic_refresh_mode_enabled && xd->segmentation_enabled)
{
const MB_MODE_INFO * mbmi = &xd->mode_info_context->mbmi;
cpi->segmentation_map[map_index + mb_col] = mbmi->segment_id;
/* If the block has been refreshed mark it as clean
* (the magnitude of the -ve influences how long it
* will be before we consider another refresh):
* Else if it was coded (last frame 0,0) and has
* not already been refreshed then mark it as a
* candidate for cleanup next time (marked 0) else
* mark it as dirty (1).
*/
if (mbmi->segment_id)
cpi->cyclic_refresh_map[map_index + mb_col] = -1;
else if ((mbmi->mode == ZEROMV) && (mbmi->ref_frame == LAST_FRAME))
{
if (cpi->cyclic_refresh_map[map_index + mb_col] == 1)
cpi->cyclic_refresh_map[map_index + mb_col] = 0;
}
else
cpi->cyclic_refresh_map[map_index + mb_col] = 1;
}
}
#if CONFIG_REALTIME_ONLY & CONFIG_ONTHEFLY_BITPACKING
/* pack tokens for this MB */
{
int tok_count = tp - tp_start;
pack_tokens(w, tp_start, tok_count);
}
#else
cpi->tplist[mb_row].stop = tp;
#endif
/* Increment pointer into gf usage flags structure. */
x->gf_active_ptr++;
/* Increment the activity mask pointers. */
x->mb_activity_ptr++;
/* adjust to the next column of macroblocks */
x->src.y_buffer += 16;
x->src.u_buffer += 8;
x->src.v_buffer += 8;
recon_yoffset += 16;
recon_uvoffset += 8;
/* Keep track of segment usage */
segment_counts[xd->mode_info_context->mbmi.segment_id]++;
/* skip to next mb */
xd->mode_info_context++;
x->partition_info++;
xd->above_context++;
}
vp8_extend_mb_row( &cm->yv12_fb[dst_fb_idx],
xd->dst.y_buffer + 16,
xd->dst.u_buffer + 8,
xd->dst.v_buffer + 8);
*current_mb_col = mb_col + nsync;
/* this is to account for the border */
xd->mode_info_context++;
x->partition_info++;
x->src.y_buffer += 16 * x->src.y_stride * (cpi->encoding_thread_count + 1) - 16 * cm->mb_cols;
x->src.u_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
x->src.v_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
xd->mode_info_context += xd->mode_info_stride * cpi->encoding_thread_count;
x->partition_info += xd->mode_info_stride * cpi->encoding_thread_count;
x->gf_active_ptr += cm->mb_cols * cpi->encoding_thread_count;
if (mb_row == cm->mb_rows - 1)
{
sem_post(&cpi->h_event_end_encoding); /* signal frame encoding end */
}
}
}
}
/* printf("exit thread %d\n", ithread); */
return 0;
}
static void decode_mb_rows(VP8D_COMP *pbi)
{
VP8_COMMON *const pc = & pbi->common;
MACROBLOCKD *const xd = & pbi->mb;
MODE_INFO *lf_mic = xd->mode_info_context;
int ibc = 0;
int num_part = 1 << pc->multi_token_partition;
int recon_yoffset, recon_uvoffset;
int mb_row, mb_col;
int mb_idx = 0;
YV12_BUFFER_CONFIG *yv12_fb_new = pbi->dec_fb_ref[INTRA_FRAME];
int recon_y_stride = yv12_fb_new->y_stride;
int recon_uv_stride = yv12_fb_new->uv_stride;
unsigned char *ref_buffer[MAX_REF_FRAMES][3];
unsigned char *dst_buffer[3];
unsigned char *lf_dst[3];
unsigned char *eb_dst[3];
int i;
int ref_fb_corrupted[MAX_REF_FRAMES];
ref_fb_corrupted[INTRA_FRAME] = 0;
for(i = 1; i < MAX_REF_FRAMES; i++)
{
YV12_BUFFER_CONFIG *this_fb = pbi->dec_fb_ref[i];
ref_buffer[i][0] = this_fb->y_buffer;
ref_buffer[i][1] = this_fb->u_buffer;
ref_buffer[i][2] = this_fb->v_buffer;
ref_fb_corrupted[i] = this_fb->corrupted;
}
/* Set up the buffer pointers */
eb_dst[0] = lf_dst[0] = dst_buffer[0] = yv12_fb_new->y_buffer;
eb_dst[1] = lf_dst[1] = dst_buffer[1] = yv12_fb_new->u_buffer;
eb_dst[2] = lf_dst[2] = dst_buffer[2] = yv12_fb_new->v_buffer;
xd->up_available = 0;
/* Initialize the loop filter for this frame. */
if(pc->filter_level)
vp8_loop_filter_frame_init(pc, xd, pc->filter_level);
vp8_setup_intra_recon_top_line(yv12_fb_new);
/* Decode the individual macro block */
for (mb_row = 0; mb_row < pc->mb_rows; mb_row++)
{
if (num_part > 1)
{
xd->current_bc = & pbi->mbc[ibc];
ibc++;
if (ibc == num_part)
ibc = 0;
}
recon_yoffset = mb_row * recon_y_stride * 16;
recon_uvoffset = mb_row * recon_uv_stride * 8;
/* reset contexts */
xd->above_context = pc->above_context;
vpx_memset(xd->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
xd->left_available = 0;
xd->mb_to_top_edge = -((mb_row * 16) << 3);
xd->mb_to_bottom_edge = ((pc->mb_rows - 1 - mb_row) * 16) << 3;
xd->recon_above[0] = dst_buffer[0] + recon_yoffset;
xd->recon_above[1] = dst_buffer[1] + recon_uvoffset;
xd->recon_above[2] = dst_buffer[2] + recon_uvoffset;
xd->recon_left[0] = xd->recon_above[0] - 1;
xd->recon_left[1] = xd->recon_above[1] - 1;
xd->recon_left[2] = xd->recon_above[2] - 1;
xd->recon_above[0] -= xd->dst.y_stride;
xd->recon_above[1] -= xd->dst.uv_stride;
xd->recon_above[2] -= xd->dst.uv_stride;
/* TODO: move to outside row loop */
xd->recon_left_stride[0] = xd->dst.y_stride;
xd->recon_left_stride[1] = xd->dst.uv_stride;
setup_intra_recon_left(xd->recon_left[0], xd->recon_left[1],
xd->recon_left[2], xd->dst.y_stride,
xd->dst.uv_stride);
for (mb_col = 0; mb_col < pc->mb_cols; mb_col++)
{
/* Distance of Mb to the various image edges.
* These are specified to 8th pel as they are always compared to values
* that are in 1/8th pel units
*/
xd->mb_to_left_edge = -((mb_col * 16) << 3);
xd->mb_to_right_edge = ((pc->mb_cols - 1 - mb_col) * 16) << 3;
#if CONFIG_ERROR_CONCEALMENT
{
int corrupt_residual = (!pbi->independent_partitions &&
pbi->frame_corrupt_residual) ||
vp8dx_bool_error(xd->current_bc);
if (pbi->ec_active &&
xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME &&
corrupt_residual)
{
/* We have an intra block with corrupt coefficients, better to
* conceal with an inter block. Interpolate MVs from neighboring
* MBs.
*
* Note that for the first mb with corrupt residual in a frame,
* we might not discover that before decoding the residual. That
* happens after this check, and therefore no inter concealment
* will be done.
*/
vp8_interpolate_motion(xd,
mb_row, mb_col,
pc->mb_rows, pc->mb_cols,
pc->mode_info_stride);
}
}
#endif
xd->dst.y_buffer = dst_buffer[0] + recon_yoffset;
xd->dst.u_buffer = dst_buffer[1] + recon_uvoffset;
xd->dst.v_buffer = dst_buffer[2] + recon_uvoffset;
if (xd->mode_info_context->mbmi.ref_frame >= LAST_FRAME) {
MV_REFERENCE_FRAME ref = xd->mode_info_context->mbmi.ref_frame;
xd->pre.y_buffer = ref_buffer[ref][0] + recon_yoffset;
xd->pre.u_buffer = ref_buffer[ref][1] + recon_uvoffset;
xd->pre.v_buffer = ref_buffer[ref][2] + recon_uvoffset;
} else {
// ref_frame is INTRA_FRAME, pre buffer should not be used.
xd->pre.y_buffer = 0;
xd->pre.u_buffer = 0;
xd->pre.v_buffer = 0;
}
/* propagate errors from reference frames */
xd->corrupted |= ref_fb_corrupted[xd->mode_info_context->mbmi.ref_frame];
decode_macroblock(pbi, xd, mb_idx);
mb_idx++;
xd->left_available = 1;
/* check if the boolean decoder has suffered an error */
xd->corrupted |= vp8dx_bool_error(xd->current_bc);
xd->recon_above[0] += 16;
xd->recon_above[1] += 8;
xd->recon_above[2] += 8;
xd->recon_left[0] += 16;
xd->recon_left[1] += 8;
xd->recon_left[2] += 8;
recon_yoffset += 16;
recon_uvoffset += 8;
++xd->mode_info_context; /* next mb */
xd->above_context++;
}
/* adjust to the next row of mbs */
vp8_extend_mb_row(yv12_fb_new, xd->dst.y_buffer + 16,
xd->dst.u_buffer + 8, xd->dst.v_buffer + 8);
++xd->mode_info_context; /* skip prediction column */
xd->up_available = 1;
if(pc->filter_level)
{
if(mb_row > 0)
{
if (pc->filter_type == NORMAL_LOOPFILTER)
vp8_loop_filter_row_normal(pc, lf_mic, mb_row-1,
recon_y_stride, recon_uv_stride,
lf_dst[0], lf_dst[1], lf_dst[2]);
else
vp8_loop_filter_row_simple(pc, lf_mic, mb_row-1,
recon_y_stride, recon_uv_stride,
lf_dst[0], lf_dst[1], lf_dst[2]);
if(mb_row > 1)
{
yv12_extend_frame_left_right_c(yv12_fb_new,
eb_dst[0],
eb_dst[1],
eb_dst[2]);
eb_dst[0] += recon_y_stride * 16;
eb_dst[1] += recon_uv_stride * 8;
eb_dst[2] += recon_uv_stride * 8;
}
lf_dst[0] += recon_y_stride * 16;
lf_dst[1] += recon_uv_stride * 8;
lf_dst[2] += recon_uv_stride * 8;
lf_mic += pc->mb_cols;
lf_mic++; /* Skip border mb */
}
}
else
{
if(mb_row > 0)
{
/**/
yv12_extend_frame_left_right_c(yv12_fb_new,
eb_dst[0],
eb_dst[1],
eb_dst[2]);
eb_dst[0] += recon_y_stride * 16;
eb_dst[1] += recon_uv_stride * 8;
eb_dst[2] += recon_uv_stride * 8;
}
}
}
if(pc->filter_level)
{
if (pc->filter_type == NORMAL_LOOPFILTER)
vp8_loop_filter_row_normal(pc, lf_mic, mb_row-1, recon_y_stride,
recon_uv_stride, lf_dst[0], lf_dst[1],
lf_dst[2]);
else
vp8_loop_filter_row_simple(pc, lf_mic, mb_row-1, recon_y_stride,
recon_uv_stride, lf_dst[0], lf_dst[1],
lf_dst[2]);
yv12_extend_frame_left_right_c(yv12_fb_new,
eb_dst[0],
eb_dst[1],
eb_dst[2]);
eb_dst[0] += recon_y_stride * 16;
eb_dst[1] += recon_uv_stride * 8;
eb_dst[2] += recon_uv_stride * 8;
}
yv12_extend_frame_left_right_c(yv12_fb_new,
eb_dst[0],
eb_dst[1],
eb_dst[2]);
yv12_extend_frame_top_c(yv12_fb_new);
yv12_extend_frame_bottom_c(yv12_fb_new);
}
static void
decode_mb_row(VP8D_COMP *pbi, VP8_COMMON *pc, int mb_row, MACROBLOCKD *xd)
{
int recon_yoffset, recon_uvoffset;
int mb_col;
int ref_fb_idx = pc->lst_fb_idx;
int dst_fb_idx = pc->new_fb_idx;
int recon_y_stride = pc->yv12_fb[ref_fb_idx].y_stride;
int recon_uv_stride = pc->yv12_fb[ref_fb_idx].uv_stride;
vpx_memset(&pc->left_context, 0, sizeof(pc->left_context));
recon_yoffset = mb_row * recon_y_stride * 16;
recon_uvoffset = mb_row * recon_uv_stride * 8;
/* reset above block coeffs */
xd->above_context = pc->above_context;
xd->up_available = (mb_row != 0);
xd->mb_to_top_edge = -((mb_row * 16)) << 3;
xd->mb_to_bottom_edge = ((pc->mb_rows - 1 - mb_row) * 16) << 3;
for (mb_col = 0; mb_col < pc->mb_cols; mb_col++)
{
/* Distance of Mb to the various image edges.
* These are specified to 8th pel as they are always compared to values
* that are in 1/8th pel units
*/
xd->mb_to_left_edge = -((mb_col * 16) << 3);
xd->mb_to_right_edge = ((pc->mb_cols - 1 - mb_col) * 16) << 3;
#if CONFIG_ERROR_CONCEALMENT
{
int corrupt_residual = (!pbi->independent_partitions &&
pbi->frame_corrupt_residual) ||
vp8dx_bool_error(xd->current_bc);
if (pbi->ec_active &&
xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME &&
corrupt_residual)
{
/* We have an intra block with corrupt coefficients, better to
* conceal with an inter block. Interpolate MVs from neighboring
* MBs.
*
* Note that for the first mb with corrupt residual in a frame,
* we might not discover that before decoding the residual. That
* happens after this check, and therefore no inter concealment
* will be done.
*/
vp8_interpolate_motion(xd,
mb_row, mb_col,
pc->mb_rows, pc->mb_cols,
pc->mode_info_stride);
}
}
#endif
xd->dst.y_buffer = pc->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
xd->dst.u_buffer = pc->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
xd->dst.v_buffer = pc->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
xd->left_available = (mb_col != 0);
/* Select the appropriate reference frame for this MB */
if (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME)
ref_fb_idx = pc->lst_fb_idx;
else if (xd->mode_info_context->mbmi.ref_frame == GOLDEN_FRAME)
ref_fb_idx = pc->gld_fb_idx;
else
ref_fb_idx = pc->alt_fb_idx;
xd->pre.y_buffer = pc->yv12_fb[ref_fb_idx].y_buffer + recon_yoffset;
xd->pre.u_buffer = pc->yv12_fb[ref_fb_idx].u_buffer + recon_uvoffset;
xd->pre.v_buffer = pc->yv12_fb[ref_fb_idx].v_buffer + recon_uvoffset;
if (xd->mode_info_context->mbmi.ref_frame != INTRA_FRAME)
{
/* propagate errors from reference frames */
xd->corrupted |= pc->yv12_fb[ref_fb_idx].corrupted;
}
decode_macroblock(pbi, xd, mb_row * pc->mb_cols + mb_col);
/* check if the boolean decoder has suffered an error */
xd->corrupted |= vp8dx_bool_error(xd->current_bc);
recon_yoffset += 16;
recon_uvoffset += 8;
++xd->mode_info_context; /* next mb */
xd->above_context++;
}
/* adjust to the next row of mbs */
vp8_extend_mb_row(
&pc->yv12_fb[dst_fb_idx],
xd->dst.y_buffer + 16, xd->dst.u_buffer + 8, xd->dst.v_buffer + 8
);
++xd->mode_info_context; /* skip prediction column */
}
static
THREAD_FUNCTION thread_encoding_proc(void *p_data)
{
#if CONFIG_MULTITHREAD
int ithread = ((ENCODETHREAD_DATA *)p_data)->ithread;
VP8_COMP *cpi = (VP8_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr1);
MB_ROW_COMP *mbri = (MB_ROW_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr2);
ENTROPY_CONTEXT_PLANES mb_row_left_context;
//printf("Started thread %d\n", ithread);
while (1)
{
if (cpi->b_multi_threaded == 0)
break;
//if(WaitForSingleObject(cpi->h_event_mbrencoding[ithread], INFINITE) == WAIT_OBJECT_0)
if (sem_wait(&cpi->h_event_mbrencoding[ithread]) == 0)
{
if (cpi->b_multi_threaded == FALSE) // we're shutting down
break;
else
{
VP8_COMMON *cm = &cpi->common;
int mb_row = mbri->mb_row;
MACROBLOCK *x = &mbri->mb;
MACROBLOCKD *xd = &x->e_mbd;
TOKENEXTRA **tp = &mbri->tp;
int *segment_counts = mbri->segment_counts;
int *totalrate = &mbri->totalrate;
{
int i;
int recon_yoffset, recon_uvoffset;
int mb_col;
int ref_fb_idx = cm->lst_fb_idx;
int dst_fb_idx = cm->new_fb_idx;
int recon_y_stride = cm->yv12_fb[ref_fb_idx].y_stride;
int recon_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride;
volatile int *last_row_current_mb_col;
if (ithread > 0)
last_row_current_mb_col = &cpi->mb_row_ei[ithread-1].current_mb_col;
else
last_row_current_mb_col = &cpi->current_mb_col_main;
// reset above block coeffs
xd->above_context = cm->above_context;
xd->left_context = &mb_row_left_context;
vp8_zero(mb_row_left_context);
xd->up_available = (mb_row != 0);
recon_yoffset = (mb_row * recon_y_stride * 16);
recon_uvoffset = (mb_row * recon_uv_stride * 8);
cpi->tplist[mb_row].start = *tp;
//printf("Thread mb_row = %d\n", mb_row);
// for each macroblock col in image
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
{
int seg_map_index = (mb_row * cm->mb_cols);
while (mb_col > (*last_row_current_mb_col - 1) && *last_row_current_mb_col != cm->mb_cols - 1)
{
x86_pause_hint();
thread_sleep(0);
}
// Distance of Mb to the various image edges.
// These specified to 8th pel as they are always compared to values that are in 1/8th pel units
xd->mb_to_left_edge = -((mb_col * 16) << 3);
xd->mb_to_right_edge = ((cm->mb_cols - 1 - mb_col) * 16) << 3;
xd->mb_to_top_edge = -((mb_row * 16) << 3);
xd->mb_to_bottom_edge = ((cm->mb_rows - 1 - mb_row) * 16) << 3;
// Set up limit values for motion vectors used to prevent them extending outside the UMV borders
x->mv_col_min = -((mb_col * 16) + (VP8BORDERINPIXELS - 16));
x->mv_col_max = ((cm->mb_cols - 1 - mb_col) * 16) + (VP8BORDERINPIXELS - 16);
x->mv_row_min = -((mb_row * 16) + (VP8BORDERINPIXELS - 16));
x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16) + (VP8BORDERINPIXELS - 16);
xd->dst.y_buffer = cm->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
xd->dst.u_buffer = cm->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
xd->dst.v_buffer = cm->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
xd->left_available = (mb_col != 0);
// Is segmentation enabled
// MB level adjutment to quantizer
if (xd->segmentation_enabled)
{
// Code to set segment id in xd->mbmi.segment_id for current MB (with range checking)
if (cpi->segmentation_map[seg_map_index+mb_col] <= 3)
xd->mode_info_context->mbmi.segment_id = cpi->segmentation_map[seg_map_index+mb_col];
else
xd->mode_info_context->mbmi.segment_id = 0;
vp8cx_mb_init_quantizer(cpi, x);
}
else
xd->mode_info_context->mbmi.segment_id = 0; // Set to Segment 0 by default
if (cm->frame_type == KEY_FRAME)
{
*totalrate += vp8cx_encode_intra_macro_block(cpi, x, tp);
#ifdef MODE_STATS
y_modes[xd->mbmi.mode] ++;
#endif
}
else
{
*totalrate += vp8cx_encode_inter_macroblock(cpi, x, tp, recon_yoffset, recon_uvoffset);
#ifdef MODE_STATS
inter_y_modes[xd->mbmi.mode] ++;
if (xd->mbmi.mode == SPLITMV)
{
int b;
for (b = 0; b < xd->mbmi.partition_count; b++)
{
inter_b_modes[x->partition->bmi[b].mode] ++;
}
}
#endif
// Count of last ref frame 0,0 useage
if ((xd->mode_info_context->mbmi.mode == ZEROMV) && (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME))
cpi->inter_zz_count ++;
}
cpi->tplist[mb_row].stop = *tp;
x->gf_active_ptr++; // Increment pointer into gf useage flags structure for next mb
for (i = 0; i < 16; i++)
vpx_memcpy(&xd->mode_info_context->bmi[i], &xd->block[i].bmi, sizeof(xd->block[i].bmi));
// adjust to the next column of macroblocks
x->src.y_buffer += 16;
x->src.u_buffer += 8;
x->src.v_buffer += 8;
recon_yoffset += 16;
recon_uvoffset += 8;
// Keep track of segment useage
segment_counts[xd->mode_info_context->mbmi.segment_id] ++;
// skip to next mb
xd->mode_info_context++;
x->partition_info++;
xd->above_context++;
cpi->mb_row_ei[ithread].current_mb_col = mb_col;
}
//extend the recon for intra prediction
vp8_extend_mb_row(
&cm->yv12_fb[dst_fb_idx],
xd->dst.y_buffer + 16,
xd->dst.u_buffer + 8,
xd->dst.v_buffer + 8);
// this is to account for the border
xd->mode_info_context++;
x->partition_info++;
x->src.y_buffer += 16 * x->src.y_stride * (cpi->encoding_thread_count + 1) - 16 * cm->mb_cols;
x->src.u_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
x->src.v_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
xd->mode_info_context += xd->mode_info_stride * cpi->encoding_thread_count;
x->partition_info += xd->mode_info_stride * cpi->encoding_thread_count;
if (ithread == (cpi->encoding_thread_count - 1) || mb_row == cm->mb_rows - 1)
{
//SetEvent(cpi->h_event_main);
sem_post(&cpi->h_event_main);
}
}
}
}
}
#else
(void) p_data;
#endif
//printf("exit thread %d\n", ithread);
return 0;
}
static
THREAD_FUNCTION thread_encoding_proc(void *p_data)
{
int ithread = ((ENCODETHREAD_DATA *)p_data)->ithread;
VP8_COMP *cpi = (VP8_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr1);
MB_ROW_COMP *mbri = (MB_ROW_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr2);
ENTROPY_CONTEXT_PLANES mb_row_left_context;
const int nsync = cpi->mt_sync_range;
//printf("Started thread %d\n", ithread);
while (1)
{
if (cpi->b_multi_threaded == 0)
break;
//if(WaitForSingleObject(cpi->h_event_mbrencoding[ithread], INFINITE) == WAIT_OBJECT_0)
if (sem_wait(&cpi->h_event_start_encoding[ithread]) == 0)
{
VP8_COMMON *cm = &cpi->common;
int mb_row;
MACROBLOCK *x = &mbri->mb;
MACROBLOCKD *xd = &x->e_mbd;
TOKENEXTRA *tp ;
int *segment_counts = mbri->segment_counts;
int *totalrate = &mbri->totalrate;
if (cpi->b_multi_threaded == 0) // we're shutting down
break;
for (mb_row = ithread + 1; mb_row < cm->mb_rows; mb_row += (cpi->encoding_thread_count + 1))
{
int recon_yoffset, recon_uvoffset;
int mb_col;
int ref_fb_idx = cm->lst_fb_idx;
int dst_fb_idx = cm->new_fb_idx;
int recon_y_stride = cm->yv12_fb[ref_fb_idx].y_stride;
int recon_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride;
int map_index = (mb_row * cm->mb_cols);
volatile int *last_row_current_mb_col;
tp = cpi->tok + (mb_row * (cm->mb_cols * 16 * 24));
last_row_current_mb_col = &cpi->mt_current_mb_col[mb_row - 1];
// reset above block coeffs
xd->above_context = cm->above_context;
xd->left_context = &mb_row_left_context;
vp8_zero(mb_row_left_context);
xd->up_available = (mb_row != 0);
recon_yoffset = (mb_row * recon_y_stride * 16);
recon_uvoffset = (mb_row * recon_uv_stride * 8);
cpi->tplist[mb_row].start = tp;
//printf("Thread mb_row = %d\n", mb_row);
// Set the mb activity pointer to the start of the row.
x->mb_activity_ptr = &cpi->mb_activity_map[map_index];
// for each macroblock col in image
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
{
if ((mb_col & (nsync - 1)) == 0)
{
while (mb_col > (*last_row_current_mb_col - nsync) && *last_row_current_mb_col != cm->mb_cols - 1)
{
x86_pause_hint();
thread_sleep(0);
}
}
// Distance of Mb to the various image edges.
// These specified to 8th pel as they are always compared to values that are in 1/8th pel units
xd->mb_to_left_edge = -((mb_col * 16) << 3);
xd->mb_to_right_edge = ((cm->mb_cols - 1 - mb_col) * 16) << 3;
xd->mb_to_top_edge = -((mb_row * 16) << 3);
xd->mb_to_bottom_edge = ((cm->mb_rows - 1 - mb_row) * 16) << 3;
// Set up limit values for motion vectors used to prevent them extending outside the UMV borders
x->mv_col_min = -((mb_col * 16) + (VP8BORDERINPIXELS - 16));
x->mv_col_max = ((cm->mb_cols - 1 - mb_col) * 16) + (VP8BORDERINPIXELS - 16);
x->mv_row_min = -((mb_row * 16) + (VP8BORDERINPIXELS - 16));
x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16) + (VP8BORDERINPIXELS - 16);
xd->dst.y_buffer = cm->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
xd->dst.u_buffer = cm->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
xd->dst.v_buffer = cm->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
xd->left_available = (mb_col != 0);
x->rddiv = cpi->RDDIV;
x->rdmult = cpi->RDMULT;
//Copy current mb to a buffer
vp8_copy_mem16x16(x->src.y_buffer, x->src.y_stride, x->thismb, 16);
if (cpi->oxcf.tuning == VP8_TUNE_SSIM)
vp8_activity_masking(cpi, x);
// Is segmentation enabled
// MB level adjutment to quantizer
if (xd->segmentation_enabled)
{
// Code to set segment id in xd->mbmi.segment_id for current MB (with range checking)
if (cpi->segmentation_map[map_index + mb_col] <= 3)
xd->mode_info_context->mbmi.segment_id = cpi->segmentation_map[map_index + mb_col];
else
xd->mode_info_context->mbmi.segment_id = 0;
vp8cx_mb_init_quantizer(cpi, x, 1);
}
else
xd->mode_info_context->mbmi.segment_id = 0; // Set to Segment 0 by default
x->active_ptr = cpi->active_map + map_index + mb_col;
if (cm->frame_type == KEY_FRAME)
{
*totalrate += vp8cx_encode_intra_macro_block(cpi, x, &tp);
#ifdef MODE_STATS
y_modes[xd->mbmi.mode] ++;
#endif
}
else
{
*totalrate += vp8cx_encode_inter_macroblock(cpi, x, &tp, recon_yoffset, recon_uvoffset);
#ifdef MODE_STATS
inter_y_modes[xd->mbmi.mode] ++;
if (xd->mbmi.mode == SPLITMV)
{
int b;
for (b = 0; b < xd->mbmi.partition_count; b++)
{
inter_b_modes[x->partition->bmi[b].mode] ++;
}
}
#endif
// Count of last ref frame 0,0 useage
if ((xd->mode_info_context->mbmi.mode == ZEROMV) && (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME))
cpi->inter_zz_count++;
// Special case code for cyclic refresh
// If cyclic update enabled then copy xd->mbmi.segment_id; (which may have been updated based on mode
// during vp8cx_encode_inter_macroblock()) back into the global sgmentation map
if (cpi->cyclic_refresh_mode_enabled && xd->segmentation_enabled)
{
const MB_MODE_INFO * mbmi = &xd->mode_info_context->mbmi;
cpi->segmentation_map[map_index + mb_col] = mbmi->segment_id;
// If the block has been refreshed mark it as clean (the magnitude of the -ve influences how long it will be before we consider another refresh):
// Else if it was coded (last frame 0,0) and has not already been refreshed then mark it as a candidate for cleanup next time (marked 0)
// else mark it as dirty (1).
if (mbmi->segment_id)
cpi->cyclic_refresh_map[map_index + mb_col] = -1;
else if ((mbmi->mode == ZEROMV) && (mbmi->ref_frame == LAST_FRAME))
{
if (cpi->cyclic_refresh_map[map_index + mb_col] == 1)
cpi->cyclic_refresh_map[map_index + mb_col] = 0;
}
else
cpi->cyclic_refresh_map[map_index + mb_col] = 1;
}
}
cpi->tplist[mb_row].stop = tp;
// Increment pointer into gf useage flags structure.
x->gf_active_ptr++;
// Increment the activity mask pointers.
x->mb_activity_ptr++;
// adjust to the next column of macroblocks
x->src.y_buffer += 16;
x->src.u_buffer += 8;
x->src.v_buffer += 8;
recon_yoffset += 16;
recon_uvoffset += 8;
// Keep track of segment useage
segment_counts[xd->mode_info_context->mbmi.segment_id]++;
// skip to next mb
xd->mode_info_context++;
x->partition_info++;
xd->above_context++;
cpi->mt_current_mb_col[mb_row] = mb_col;
}
//extend the recon for intra prediction
vp8_extend_mb_row(
&cm->yv12_fb[dst_fb_idx],
xd->dst.y_buffer + 16,
xd->dst.u_buffer + 8,
xd->dst.v_buffer + 8);
// this is to account for the border
xd->mode_info_context++;
x->partition_info++;
x->src.y_buffer += 16 * x->src.y_stride * (cpi->encoding_thread_count + 1) - 16 * cm->mb_cols;
x->src.u_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
x->src.v_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
xd->mode_info_context += xd->mode_info_stride * cpi->encoding_thread_count;
x->partition_info += xd->mode_info_stride * cpi->encoding_thread_count;
x->gf_active_ptr += cm->mb_cols * cpi->encoding_thread_count;
if (mb_row == cm->mb_rows - 1)
{
//SetEvent(cpi->h_event_main);
sem_post(&cpi->h_event_end_encoding); /* signal frame encoding end */
}
}
}
}
//printf("exit thread %d\n", ithread);
return 0;
}
static void decode_mb_rows(VP8D_COMP *pbi)
{
VP8_COMMON *const pc = & pbi->common;
MACROBLOCKD *const xd = & pbi->mb;
int ibc = 0;
int num_part = 1 << pc->multi_token_partition;
int recon_yoffset, recon_uvoffset;
int mb_row, mb_col;
int mb_idx = 0;
int dst_fb_idx = pc->new_fb_idx;
int recon_y_stride = pc->yv12_fb[dst_fb_idx].y_stride;
int recon_uv_stride = pc->yv12_fb[dst_fb_idx].uv_stride;
unsigned char *ref_buffer[MAX_REF_FRAMES][3];
unsigned char *dst_buffer[3];
int i;
int ref_fb_index[MAX_REF_FRAMES];
int ref_fb_corrupted[MAX_REF_FRAMES];
ref_fb_corrupted[INTRA_FRAME] = 0;
ref_fb_index[LAST_FRAME] = pc->lst_fb_idx;
ref_fb_index[GOLDEN_FRAME] = pc->gld_fb_idx;
ref_fb_index[ALTREF_FRAME] = pc->alt_fb_idx;
for(i = 1; i < MAX_REF_FRAMES; i++)
{
ref_buffer[i][0] = pc->yv12_fb[ref_fb_index[i]].y_buffer;
ref_buffer[i][1] = pc->yv12_fb[ref_fb_index[i]].u_buffer;
ref_buffer[i][2] = pc->yv12_fb[ref_fb_index[i]].v_buffer;
ref_fb_corrupted[i] = pc->yv12_fb[ref_fb_index[i]].corrupted;
}
dst_buffer[0] = pc->yv12_fb[dst_fb_idx].y_buffer;
dst_buffer[1] = pc->yv12_fb[dst_fb_idx].u_buffer;
dst_buffer[2] = pc->yv12_fb[dst_fb_idx].v_buffer;
xd->up_available = 0;
/* Decode the individual macro block */
for (mb_row = 0; mb_row < pc->mb_rows; mb_row++)
{
if (num_part > 1)
{
xd->current_bc = & pbi->mbc[ibc];
ibc++;
if (ibc == num_part)
ibc = 0;
}
recon_yoffset = mb_row * recon_y_stride * 16;
recon_uvoffset = mb_row * recon_uv_stride * 8;
/* reset contexts */
xd->above_context = pc->above_context;
vpx_memset(xd->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
xd->left_available = 0;
xd->mb_to_top_edge = -((mb_row * 16)) << 3;
xd->mb_to_bottom_edge = ((pc->mb_rows - 1 - mb_row) * 16) << 3;
xd->recon_above[0] = dst_buffer[0] + recon_yoffset;
xd->recon_above[1] = dst_buffer[1] + recon_uvoffset;
xd->recon_above[2] = dst_buffer[2] + recon_uvoffset;
xd->recon_left[0] = xd->recon_above[0] - 1;
xd->recon_left[1] = xd->recon_above[1] - 1;
xd->recon_left[2] = xd->recon_above[2] - 1;
xd->recon_above[0] -= xd->dst.y_stride;
xd->recon_above[1] -= xd->dst.uv_stride;
xd->recon_above[2] -= xd->dst.uv_stride;
//TODO: move to outside row loop
xd->recon_left_stride[0] = xd->dst.y_stride;
xd->recon_left_stride[1] = xd->dst.uv_stride;
for (mb_col = 0; mb_col < pc->mb_cols; mb_col++)
{
/* Distance of Mb to the various image edges.
* These are specified to 8th pel as they are always compared to values
* that are in 1/8th pel units
*/
xd->mb_to_left_edge = -((mb_col * 16) << 3);
xd->mb_to_right_edge = ((pc->mb_cols - 1 - mb_col) * 16) << 3;
#if CONFIG_ERROR_CONCEALMENT
{
int corrupt_residual = (!pbi->independent_partitions &&
pbi->frame_corrupt_residual) ||
vp8dx_bool_error(xd->current_bc);
if (pbi->ec_active &&
xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME &&
corrupt_residual)
{
/* We have an intra block with corrupt coefficients, better to
* conceal with an inter block. Interpolate MVs from neighboring
* MBs.
*
* Note that for the first mb with corrupt residual in a frame,
* we might not discover that before decoding the residual. That
* happens after this check, and therefore no inter concealment
* will be done.
*/
vp8_interpolate_motion(xd,
mb_row, mb_col,
pc->mb_rows, pc->mb_cols,
pc->mode_info_stride);
}
}
#endif
xd->dst.y_buffer = dst_buffer[0] + recon_yoffset;
xd->dst.u_buffer = dst_buffer[1] + recon_uvoffset;
xd->dst.v_buffer = dst_buffer[2] + recon_uvoffset;
xd->pre.y_buffer = ref_buffer[xd->mode_info_context->mbmi.ref_frame][0] + recon_yoffset;
xd->pre.u_buffer = ref_buffer[xd->mode_info_context->mbmi.ref_frame][1] + recon_uvoffset;
xd->pre.v_buffer = ref_buffer[xd->mode_info_context->mbmi.ref_frame][2] + recon_uvoffset;
/* propagate errors from reference frames */
xd->corrupted |= ref_fb_corrupted[xd->mode_info_context->mbmi.ref_frame];
decode_macroblock(pbi, xd, mb_idx);
mb_idx++;
xd->left_available = 1;
/* check if the boolean decoder has suffered an error */
xd->corrupted |= vp8dx_bool_error(xd->current_bc);
xd->recon_above[0] += 16;
xd->recon_above[1] += 8;
xd->recon_above[2] += 8;
xd->recon_left[0] += 16;
xd->recon_left[1] += 8;
xd->recon_left[2] += 8;
recon_yoffset += 16;
recon_uvoffset += 8;
++xd->mode_info_context; /* next mb */
xd->above_context++;
}
/* adjust to the next row of mbs */
vp8_extend_mb_row(
&pc->yv12_fb[dst_fb_idx],
xd->dst.y_buffer + 16, xd->dst.u_buffer + 8, xd->dst.v_buffer + 8
);
++xd->mode_info_context; /* skip prediction column */
xd->up_available = 1;
}
}